Phil 6.22.2022

I think this is one of those interesting posts about how AI is a tool like other tools. It’s a valid point, but I’m not so sure. In my creative experience there is an initial creative part and then an extensive editing part. Generating that initial content is hard in cases like writing, graphic arts, chorography, etc. It’s not as hard when working with found objects (like photography or this piece by Marcel Duchamp. AI models like Dall-e and GPT-3 change this balance and make the initial creation more working from found objects that are latent in the models.
Look! Trajectories!

Online Coordination: Methods and Comparative Case Studies of Coordinated Groups across Four Events in the United States

  • Coordinated groups of user accounts working together in online social media can be used to manipulate the online discourse and thus is an important area of study. In this study, we work towards a general theory of coordination. There are many ways to coordinate groups online: semantic, social, referral and many more. Each represents a coordination dimension, where the more dimensions of coordination are present for one event, the stronger the coordination present. We build on existing approaches that detect coordinated groups by identifying high levels of synchronized actions within a specified time window. A key concern with this approach is the selection of the time window. We propose a method that selects the optimal window size to accurately capture local coordination while avoiding the capture of coincidental synchronicity. With this enhanced method of coordination detection, we perform a comparative study across four events: US Elections Primaries 2020, Reopen America 2020, Capitol Riots 2021 and COVID Vaccine Release 2021. Herein, we explore the following three dimensions of coordination for each event — semantic, referral and social coordination — and perform group and user analysis within and among the events. This allows us to expose different user coordination behavior patterns and identify narratives and user support themes, hence estimating the degree and theme of coordination.

Pinged Nabeel. We’ll see where that goes

Sent intro to Shannon for Aaron – nope, can’t test out of a BS?


  • Send a note to MIT Press and see if I can get a proposal template – done!

GPT Agents

  • Test some downloads
  • Create experiment, query, tweet database (users table later, based on SELECT DISTINCT on user id’s from the tweet table) got the experiment and query DBs working. Tweets are harder
  • Coding
    • Remove Sample Multiple – done
    • Remove Balance (Individual == balanced. Maybe rename?) -done
    • Create a unique experiment name (each time the App is launched?) – done
  • See if I can have a chat with Aaron about clustering


  • Set up a file for papers/websites/repos of monolithic models and start looking
  • Set up an Overleaf doc for the survey?